| name | questrade-refresh-prices |
| description | Refreshes live market prices into domain_model.sqlite via Questrade get_quotes — current holdings by default, or the full watchlist with --full-watchlist. |
| argument-hint | [--dry-run] [--full-watchlist] |
| allowed-tools | Bash, Read |
Refresh Prices
Purpose
Fetches live quotes for USD-denominated, non-cash investments via Questrade's get_quotes MCP tool and writes them into investment_price — an optional, user-triggered augment to the existing TradingView/yfinance pricing baseline (Rule #20), fully separate from questrade-sync-portfolio's holdings/balances sync.
Prerequisites & Pre-Flight Check
- Verify Questrade MCP session is active via
List Accounts.
- If unauthenticated, prompt user to run
/questrade:questrade-setup (/mcp -> questrade -> Log in).
Schema Reference
See references/questrade-tool-schemas.md (get_quotes section) for exact tool param names and response shapes — do not re-derive or guess these from memory.
Skill-Specific Behavior
- Never runs automatically. This skill is a fully separate, user-triggered action from
questrade-sync-portfolio — a holdings sync never implicitly refreshes prices, and this skill never implicitly re-syncs holdings.
- Excludes non-tradeable cash rows (
CASH_USD, asset_class == "CASH") from any quote request — get_quotes would never resolve a synthetic symbol like that anyway.
- Excludes non-USD-denominated investments (e.g.
DLR.TO, PSU-U.TO). The LIVE quote's own currency field is the authoritative check — not the stored investment.currency column, which is unreliable (this domain model hardcodes currency='USD' on every investment row today, including genuinely CAD-denominated tickers like PSU-U.TO). questrade_price_refresh.py's persist_quotes_to_prices() checks quote["currency"], not the stored value, so a mislabeled ticker's price is still caught and skipped rather than silently written under the wrong currency.
- Batches at 20 symbols per
get_quotes call (Questrade's per-call cap).
- Two scopes: default is currently-held positions only (
account_investment.quantity > 0); --full-watchlist widens this to every investment row with is_watchlisted = 1 regardless of holding — _select_investments_for_quote_refresh() itself already queries the whole investment table (not joined to holdings), so no code change is needed to support this, only which symbol list the skill stages. A full watchlist run is typically 60-70+ symbols → 4 get_quotes calls.
Workflow
- Determine eligible symbols — pick the scope:
- Default (holdings only):
SELECT DISTINCT i.symbol FROM investment i JOIN account_investment ai ON ai.investment_id = i.investment_id WHERE i.asset_class != 'CASH' AND i.currency = 'USD' AND ai.quantity > 0
--full-watchlist: SELECT symbol FROM investment WHERE asset_class != 'CASH' AND currency = 'USD' AND is_watchlisted = 1
- Batch and fetch quotes: split the symbol list into groups of ≤20, call MCP tool
Get Quotes(symbols=[...]) once per batch.
- Stage the payload: merge all batch responses into one JSON file at
temp/questrade_price_refresh_payload.json:
{ "quotes": { "BTDR": { "symbol": "BTDR", "currency": "USD", "lastPrice": 11.845, ... }, ... } }
- Execute the persistence script:
python3 plugins/questrade/skills/questrade-refresh-prices/scripts/questrade_price_refresh.py --payload temp/questrade_price_refresh_payload.json
- Clean up & report: remove the temporary payload file, then display a summary — symbols updated, symbols skipped (and why: non-USD currency, CASH asset_class, or no usable price in the quote response).
Continuous Self-Evolution Policy
Per .agent/rules/self-evolution-policy.md:
Whenever actual MCP tool schema responses reveal unexpected parameter names, response fields, or missing attributes during live execution, agents MUST immediately refine references/questrade-tool-schemas.md (the canonical, hub-and-spoke source) first — not this file — then this file's ## Skill-Specific Behavior section only for details unique to this skill's workflow.
